Senior LLM Researcher - Post-Training & Alignment (Remote)

techire ai; San Francisco (CA); On-site; USD 350,000 - 500,000; Full time

Benefits offered by this job

Stock options

Remote work worldwide

Competitive compensation

Job summary

techire ai is a Series A company building the next generation of conversational AI, with an ex-NVIDIA & Meta research leader and a co-creator behind open-source models. The team powers hundreds of millions of conversations monthly, offering work on research that scales to real-world deployments.

They're seeking researchers who have owned meaningful post-training improvements and can tackle data, modelling, evaluation and RL infrastructure challenges to raise model quality and alignment at scale.

Qualifications

• Experience leading post-training projects delivering tangible capability improvements.
• Understanding practical challenges behind post-training techniques, not just theory.
• Experience across data, modelling, evaluation and RL infrastructure, improving model quality.

Responsibilities

• Design and run post-training experiments using techniques like DPO, GRPO, SFT and rejection sampling.
• Build and scale reinforcement learning and post-training infrastructure.
• Develop reward models, evaluation frameworks and assessment rubrics to improve model quality and behavior.
• Collaborate with vendors and data partners to curate high-quality post-training datasets and feedback pipelines.
• Own end-to-end projects spanning data collection, modelling, evaluation and infrastructure, driving improvements in reasoning and alignment.

Skills

Post-training projects

RL infrastructure

Evaluation frameworks

Data partnerships

End-to-end project ownership

Job description

techire ai is a Series A company building the next generation of conversational AI, with an ex-NVIDIA & Meta research leader and a co-creator behind open-source models. The team powers hundreds of millions of conversations monthly, offering work on research that scales to real-world deployments.

They're seeking researchers who have owned meaningful post-training improvements and can tackle data, modelling, evaluation and RL infrastructure challenges to raise model quality and alignment at scale.